The crawl script that comes with all the version of Nutch I have checked set the local/distributed operating mode using a relative path (i.e. "../*nutch-*.job").
Bash seems to be taking this as relative to the location that the crawl script was called from, not the scripts actual location.
The result is that the script thinks it is in local mode because it cannot find the job file. When trying to carry out a crawl jobs are submitted to Hadoop properly, but ifs that test for local (or not) mode fail and give strange results/result in crashes.
Using the first bash snippet from here I have modified the crawl script to look for a job file relative to the script location on disk.
I have attached a patch with my modifications.